your premise is flawed. using AI to do this shit is like using a twin-engine F-14 Tomcat’s afterburner to clear the snow from your driveway. and to get you started to ditch your shovels and shit and help your muscles atrophy, Northrop-Grumman is subsidising the insanely humongous fuel bill and maintenance for the first year.

it’s not sustainable. it’s not ethical. it’s fucking disastrous to the environment, local and at large. and it tends to blow up the house you’ve built, for no fucking reason.

I tried use AI to help me with self host. It gave me incorrect information and I wasted hours trying to read the docs and the AI slop to make sense of everything.

In the end, reading the docs was all that I needed.

I can see some value using AI. I treat it like a search engine, since Google, DDG, Bing, etc all seem to suck now favoring ads and purchased result slots rather than finding what you’re looking for. Using AI helps since it’s doing several searches for you. But as others have said, the results can be flawed, particularly when searching for niche topics.

I have zero coding experience and used ChatGPT (I think) to vibe code a simple battery GUI for my Thinkpad running linux since the OS version I had would not display the internal and external batteries separately. Technically it worked, but it looked and ran like ass.

Ultimately I think it falls apart when things go wrong and trying to coax the robot to help you get your way out of trouble, which can be very problematic.

This is such a broad topic, that we need some qualifiers to focus. Ai tools in servers from big companies is an absolute no. I am experimenting for fun with local tools on my hardware only. Then there is a huge difference in generative Ai tools that generate code, assist tools to in example translate human languages, or if you use it to ask questions in example. So there is a huge difference in how, for what, and what you actually use Ai.

Ai tools can be useful, under the right circumstances and conditions. But it comes with the problem that they mostly have scraped data without consent and ignoring the licenses. And for server based Ai you also give up your privacy and you need internet connection at all times; loosing control. Then there is this generating code thing. If you don’t program yourself and don’t understand and generate lots of code, then it will be spaghetti code, with hallucinations and you don’t understand every detail to check it.

You can AI as hard as you like, you’ll probably not manage to replace the really complex tools.

I have just started writing a laser cutter tool, and I tried some suggestions from AI for doing basic matrix transform calculations, and it was completely wrong. So I wouldn’t trust it for anything moderately complex. That said, it was fantastic for the UI.

Things are moving so fast that anything that doesn’t work today, will work tomorrow, or ‘next week’. Take all the failed attempts from others, as possibilities of what you will be able to do in a week, and catalog what you think could be beneficial for you later. They are all on the plate.

AI is a technically a cognitive extension much like pen/paper - we are offloading cognition externally. So everything you spend mental energy on now, can and will be lessened when AI takes some of your cognitive load off.

Ignore their current baby-step errors. Think about where you spend your thoughts - your cognitive energy, and think of a way an AI could assist you with that. Worry too much ? Need a push/nudge with social connections ? Tired of seeking bargains at shops that cheat you ? Have something you find difficult ?

Then imagine what you can do with like-minded people from all over the world.